Talanta 67 (2005) 908–917

Single-crystal sapphire-fiber optic sensors based on surface plasmonresonance spectroscopy for in situ monitoring

Yoon-Chang Kim, Jean-Francois Masson, Karl S. Booksh∗

Department of Chemistry and Biochemistry, Arizona State University, Tempe, AZ 85287-1604, USA

Received 22 February 2005; received in revised form 1 April 2005; accepted 4 April 2005Available online 24 August 2005

Abstract

Single-crystal sapphire-fiber optic sensors based on surface plasmon resonance (SPR) for refractive index (RI) measurements of aqueousand hydrothermal water solutions are described. Accurate measurement of RIs is essential to efficient operation and control of broad rangeof engineering processes. Some of these processes are carried out with harsh environments, such as high-temperature, high pressure, andchemical corrosion. These extreme physical conditions are proving a limiting factor in application of the conventional silica-based opticalsensors. Single-crystal sapphire is an ideal material for sensor applications, where reliable performance is required in the extreme environmentc tive indexo R sensor.H queouss angle pairw©

K

1

(bSbowbpgsts

nlyandm-the

i-redptd.usedsed

l orby

useini-msion

tly

0d

onditions. With regard to the liquid species detection, most applications of SPR sensors are designed to function near the refracf water (1.3330 RI). The RI changes of aqueous solution can be easily monitored by silica-fiber (RI, 1.4601 at 550 nm) based SPowever, the sapphire waveguide has a prohibitively high RI (1.7708 at 546 nm) for unmodified monitoring of the RI changes of aolutions. For that purpose, a practical SPR probe geometry has been applied to the ability to tune the SPR coupling wavelength/ith sapphire-fiber based SPR probe.2005 Published by Elsevier B.V.

eywords: Fiber optic sensor; Surface plasmon resonance; Sapphire-fiber; Rapid thermal annealing

. Introduction

The optical phenomenon of surface plasmon resonanceSPR) has many applications ranging from the analysis ofiomaterials to environmental inspection. The technique ofPR is based on an electromagnetic phenomenon that is capa-le of monitoring refractive index (RI)[1,2] and thicknessf the dielectric medium[3,4] that come into near contactith sensing surface. SPR sensing platforms are traditionallyased on the Kretschmann configuration[5,6], consisting of arism with thin layer of highly conductive metal (most oftenold or silver) deposited onto one face to form the sensingurface. The reflection characteristics of a medium adjacento the metal film are highly sensitive to RI variations in theample medium.

∗ Corresponding author. Tel.: +1 480 965 3058; fax: +1 480 965 2747.E-mail address: booksh@asu.edu (K.S. Booksh).

Two main types of the SPR sensors have commobeen employed: constant-angle SPR (i.e., spectral SPR)constant-wavelength SPR (i.e., angular SPR). In the prisbased SPR sensor, the surface plasmon wave (SPW) onprism can be excited by light with a modulated angle of incdence. Otherwise, the reflected intensity can be measuby wavelength modulation while the incident angle is kefixed and the wavelength of the incident light is changeTherefore the prism-based SPR sensor system can beeither as a spectral or angular SPR sensor. The prism-baSPR sensors are not, however, optimal for in situ industriaenvironmental process monitoring. The sensing region isnecessity more bulky that single fiber optics probe becaof the rigid geometry of the sensor head. Thus, even the mSpreeta by Texas Instruments requires fluid handling systeto bring the sample to the sensor head instead of immersof the sensor in the sample.

The use of optical fibers for the SPR sensing has recenattracted much attention[1,7–18] because the optical fiber

039-9140/$ – see front matter © 2005 Published by Elsevier B.V.oi:10.1016/j.talanta.2005.04.022

Y.-C. Kim et al. / Talanta 67 (2005) 908–917 909

sensors have several distinct advantages over the prism-based sensor. They are fundamentally simple in structure,less costly, require smaller sample volumes, and are amenabletowards remote sensing applications. In the traditional fiber-based SPR sensor system, there is not a fixed incident anglebut a range of angles that are allowed to propagate in themultimode fiber probe. The incident light consists of a widerange of wavelengths launched into the fiber in order to excitethe SPW across a range of angles defined by the numericalaperature of the fiber optic waveguide. However, the SPRcoupling wavelength can be tuned selectively by modifyingthe geometry of the probe tip[1,15–17]. The probe geome-try allows the sensor to be easily placed into the sample ofinterest, and the orientation of the sensing region requires aminimal sample volume, making significant invasion of thesample unnecessary[17].

Development of fiber optic SPR sensors for RI moni-toring applications where the RI of the media is far fromthe RI of the waveguide is not trivial. Most SPR sensorsare designed with incident angles in the range of 70◦–89◦from normal to monitor samples within 0.1 refractive unitsof the waveguide. BK7, SF10, or silica substrates (RI∼ 1.42)are employed for aqueous applications (RI∼ 1.34). Straightsapphire rod (RI∼ 1.77) has been employed for SPR sen-sors, but the application was the monitoring of epoxy curingprocesses (RI > 1.48)[18]. Such a range of applications trans-l e isw thed sulti real-i than0 le ofi lyc omt SPRf r thec geta olli-m ngleso bersm per-a ers.E NA)t tt roada RI oft nsorp

re theR theRp solu-t igno undd ts, or“ cally

active areas of seafloor, at an average depth of∼2100 m andin the both the Atlantic and Pacific oceans. In these regions,seawater seeps into cracks in the sea floor. This seawater isheated by the magma that lies beneath the earth’s crust, andleaches minerals from the magmatic basalts. After a com-plex circulation, the hydrothermal fluids are emitted into thecold deep-sea water. These hydrothermal fluids are acidic,reduced and enriched with chemicals including heavy met-als, methane and hydrogen sulphide. These near supercriticalconditions of sulphide mineral-enriched water corrode mostinstrument platforms; sapphire, gold, diamond, and titaniumare four of the few materials stable in these environments. Todate, sensors for temperature, pH, H2, and sulphide concen-tration have been the only successfully deployed instrumentsin hydrothermal vents[19–21]. The sapphire-fiber basedSPR probe after the RTA treatment is highly desirable in awide range of the engineering applications with harsh envi-ronments such as the natural seafloor hydrothermal vents,high-temperature combustion, and industrial processes mon-itoring of corrosive reagents. Additionally, the system couldbe applied to the development of advanced high-temperaturematerials.

2. Experimental

2

ribed[ SPRp -i lightf Poly-i2 e-t mperi 00 ho r thes berw thei rnedl pecs d ona . Fort 70 Ms rrowt with1

besw bersw -t theb bers( vea NAw bs)

ate well to fiber optic sensors, the angle of incidencithin the numerical aperature of the optical fiber andistribution of angles propagated by the fiber do not re

n an overly broad SPR spectral feature. However, toze the sensitivity to refractive index changes greater.1 refractive units less than the waveguide, the ang

ncidence must approach 45◦. Here two problems mutualompound to prohibit direct use of optical fibers. First, frhe Fresnel equations, the angular sensitivity (change ineature location with respect to change in angle) is greateloser to normal the angle of incident becomes; thus tosharp spectral feature, strict control of incident light cation must be realized. Second, to realize the acute af incidence in optical fibers, high numerical aperature fiust be employed. However, the larger the numerical ature results in a broader distribution of angle in the fibven moderate numerical aperature fibers (0.22 or 0.39

apered, in the manner of Obando and Booksh[1], to presenhe appropriate angle of incidence yield SPR features so bs to be useless when the RI of the media is far from the

he waveguide without redesign of the fiber optic SPR selatform.

Presented here are SPR monitoring applications wheI of the media being monitored is significantly less thanI of the waveguide. Single-crystal sapphire (RI∼ 1.77) SPRrobes are employed to monitor the density of aqueous

ions (RI∼ 1.33). Here the ultimate application is the desf a fiber optic sensor to monitor fluid density in and aroeep-sea hydrothermal vents. Seafloor hydrothermal venblack smokers”, have been discovered in many tectoni

.1. Probe designs

The fiber optic SPR system has been previously desc1]. The sensor system associated with the fiber opticrobe is shown inFig. 1-1. A bifurcated 0.22± 0.02 numer

cal aperture optical fiber jumper was used to guide therom the source, to the probe, and back to the detector.mide clad fibers (Polymicro Technologies) having 200± 8,20± 5 and 240± 5�m core, cladding and buffer diam

ers, respectively, were employed to manufacture the jun house. A white light LED that can be operated over 1n a commercial 9 V household battery was employed foource. Light from the LED was focused on the optical fiith a SMA fiber optic coupling lens (Thor Labs). For

nitial tests and the analyses with silica SPR probes, retuight from 502 to 950 nm was dispersed by a Kaiser Hollospectrograph (Kaiser Optical Systems Inc.) and collectePrinceton Instruments CCD camera (Roper Scientific)

he analyses with sapphire SPR probes, a Jobin-SPEX 2pectrometer with 1800 gr/mm grating was used to nahe spectral range to 42.8 nm. The spectra were collected024× 256 pixels Andor CCD camera.

Three different types of optical fibers for SPR proere employed in these experiments. All three optical fiere selected to be approximately 400�m in core diame

er to maximize the coupling between the probe tip andifurcated fiber jumper. Single-crystal sapphire optical fiPhotran, LLC) were 425�m in diameter and did not hacladding or buffer. Silica core optical fibers with 0.39ere employed. The 0.39 NA multimode fibers (Thor la

910 Y.-C. Kim et al. / Talanta 67 (2005) 908–917

Fig. 1. Schematic diagram of experimental set-up and probe designs: (1-2a) fiber based straight SPR probe; (1-2b) fiber based tapered SPR probe. (1)spectrometer; (2) white light source; (3) SPR probe; (4) sensing surface; (5) mirror.

had 400± 8, 425± 1 and 730± 3�m core, cladding andbuffer diameters, respectively.

Two probe designs were selected for presentation in thispaper (Fig. 1-2). In traditional straight SPR probe (Fig. 1-2a),there is not a fixed angle of incidence, but rather a range ofincident angles that are allowed to propagate in the straightSPR probe. The dual-tapered SPR probe (Fig. 1-2b) has sens-ing surface and the front mirror polished at orthogonal anglesbetween the two surfaces. The white light from the illumi-nating fiber/mirror reflects off of the tapered surfaces of theprobe and is incident onto the sensing surface at the designedangle,θinc. The preparation of the sensors has been exten-sively described[1]. In summary the buffer and claddingare removed from the silica-fibers; the sapphire-fibers hadno buffer or cladding. The fiber optic tip was polished to adesired angle with a fiber polisher (Ultra Tec). The surfacewas polished with 6, 3, 1 and 0.5�m lapping films. In a mod-ified Cressington 204 h sputter coater (Cressington), 1 nm ofCr followed by 50 nm of Au were deposited on the SPR activeregions. For the rapid thermal annealing (RTA), the sapphireSPR probe was placed in a fused silica tube. After the fusedsilica tube was evacuated and flame-sealed, it was heated at1200± 50◦C for 30 s.

2.2. Experimental design

bym tion.R ital-rG

nd/orpp ula-t nts.T mL,a iston

pump (Eldex Laboratories), connected to a sample reser-voir. Pressure was read from an analogue gauge incrementedat 50 psi. Heating was accomplished using flexible electricheating tape (Barnstead/Thermolyne) with variable-voltagecontroller. The temperature was read from an internal K-typethermocouple attached to a digital thermometer (Omega),accurate to±0.5◦C. For mounting the SPR probe into SMA,epoxy (Tra-Con, TRA-BOND F202) was generously placedon the SPR probe and SMA. This epoxy was also used tomount two fibers (illuminating and collecting optical fibers)into one steel jacket for manufacturing a bifurcated fiber opticcoupler that was terminated with SMA connector. The SPRprobes were linked to the coupler that was kept in alignmentwith the optic of the system.

For the analysis, an air reflectance spectrum was employedfor the reflectance reference. Each collected spectrum wasnormalized to fraction reflectance across all wavelengths rel-ative to the reference spectrum. A unique reference spectrumwas collected for every sensor studied.

2.3. Data analysis

Reflectivity references were collected before each set ofsamples with the fiber optic sensors. References were col-lected in media where the SPR response would be far outsideo thea

e the-o aru[ set samemt theo thet ationa woulda uares

Sixteen KCl (Fisher Scientific) solutions from 0 to 20%ass were prepared by dilutions from a stock KCl soluIs of the samples were verified by an automatic dig

efractometer accurate to 0.00001 RI (A. Kruss OptronicmbH).A stainless steel chamber for the high-temperature a

ressure conditions was constructed using Swagelok® com-onents. The chamber was designed to allow the sim

ion of liquid sample in the seafloor hydrothermal vehe total volume of chamber was approximately 10nd pressure was attained using an HPLC sapphire p

f the spectral window. Normalizing each spectrum byppropriate reference yields the familiar SPR dip.

Observed SPR spectral features were compared to thretical SPR features from the model proposed by Ishim

22]. A Matlab (Mathworks) program was written in houo calculate the theoretical SPR response. This is theodel as was employed by Johnston et al.[23]. Note that the

heoretical SPR model assumed TM polarized light whileptical fiber probes use un-polarized light. Thus, where

heoretical spectral profiles approach complete attenut the spectral feature, the observed spectral featurespproach 50% attenuation. The non-negative least sq

Y.-C. Kim et al. / Talanta 67 (2005) 908–917 911

fitting algorithm in Matlab was employed to fit theoreticalSPR spectra to the observed SPR spectra. The fact that SPRspectra are actually attenuation from unity, not additive froma zero baseline like absorbance spectra, was accounted for inthe assumed linear additive spectral mixing model.

3. Results and discussion

3.1. Limits of straight fiber optic SPR probes

SPR spectral features are readily observed in aqueoussolutions using a straight silica-fiber with the cladding andbuffer removed[1]. In this most recent work, 0.39 NATECSTM clad multimode silica-fibers were used for man-ufacturing the straight SPR probes. The SPW is excited atmany angles simultaneously and the SPR spectrum is accu-mulated spectra of the entire range of the propagating angles,from 70◦ to 89◦. Thus the observed SPR spectral feature isbroader than commonly seen with single angle prism-basedSPR systems.

The SPR dip is not observed with the straight core-onlysapphire probes. The sapphire core has a refractive index∼1.76; thus the theoretical NA of the fiber in aqueous solu-tion is approximately 1.45. Therefore, this would suggest thata ans-m r iss ghtsr di cha ans-mm re is0 opa-g ta does

not fall within the spectral range covered by the system used.Fig. 2b illustrates the normalized SPR spectrum of 10% KClsolution in straight sapphire-fiber based SPR probe. The theo-retical (Fig. 2a) and experimental (Fig. 2b) spectra are in goodagreement with one another. As a result, the SPW could notbe produced on the range of internal incident angles, 83◦–89◦in sapphire-fiber SPR probe. The attenuation observed below600 nm is non-SP based attenuation predicted by the Fresnelequations.

3.2. Sapphire tapered fiber optic SPR probes foraqueous phase RI monitoring

The incident angle of tapered sapphire-fiber optic probeis optimized in design for exciting SPR. The light reflects offthe tapered side at the designed angle and is directed back inparallel to the incident light through the mirror. The light fromthe mirror is also reflected off the tapered side with the sameincident angle. The sensing surface was prepared by succes-sive vacuum deposition of chromium (thickness, 1 nm) andgold (thickness, 50 nm) after thoroughly cleaning the fiber.The mirror was fabricated with reflective metals, chromiumand gold deposited onto the region with such thickness greaterthan 200 nm so as not to excite SPR.

Fig. 3b illustrates the measured SPR spectral features oft ◦ ◦ ◦ ◦ ◦ ◦( ls ell,b theirt g ina akea I ofa thea nor-mi h them robes

F 0% KC( nm of

ny angle on the input face of sapphire-fiber would be tritted. However the theoretical NA of a sapphire-fibe

ignificantly larger than the functional NA because of licattering from defects on large acceptant angles[24]. As aesult an effective numerical aperture, NAeff has been applien this case. The NAeff is defined as the sine of the launngle at which the transmission drops to 50% of the trission with a 0◦ of launch angle. In this study, the NAeff ofultimode sapphire-fiber used for SPR probe manufactu.12. Therefore, the sapphire-fiber supports internal prating angles of light from 83◦ to 90◦. Within this incidenngular region, wavelength required to excite the SPW

ig. 2. SPR spectra from straight sapphire-fiber SPR probe. Medium, 13) θinc = 88◦; (4) θinc = 89◦; Kaiser Hollospec spectrograph with 449.61

hree dual-tapered probes with 57/33 , 59 /31 , and 61/29i.e., θinc @ sensing surface/θinc @ mirror) in a 5 wt.% KColution. The light is incident upon the mirrors as wut SPR on the mirrors is not activated because ofhick thickness. The distribution of angles propagatin

sapphire-fiber has a mean too far from normal to mn effective fiber optic SPR sensor for monitoring the Rqueous solutions without tapering the fiber tip to bringngle of incidence at the SPR sensing region nearer toal. The Fresnel equations predict that a 61◦–57◦ angle of

ncidence is required to observe the SPR attenuation witaterials employed to construct the sapphire based p

l solution in water (RI, 1.35148 at room temperature): (1)θinc = 86◦; (2)θinc = 87◦;spectral coverage; the atmosphere and room temperature.

912 Y.-C. Kim et al. / Talanta 67 (2005) 908–917

Fig. 3. SPR spectra from tapered sapphire-fiber SPR probe (three angles and one RI). Medium, 5% KCl solution in water (RI, 1.34570 at room temperature):(1) θinc = 57◦; (2) θinc = 59◦; (3) θinc = 61◦; Kaiser Hollospec spectrograph with 449.61 nm of spectral coverage; the atmosphere and room temperature.

(Fig. 3a). Tapering the sapphire-fibers such that the angle ofincidence on the tip is between 57◦ and 61◦ does return auseful SPR spectral feature in qualitative agreement with thetheoretical SPR spectrum. However, the observed spectralfeature is much broader than the theoretical spectrum due tothe fact that the wavelength of maximum attenuation is verysensitive to imperfections in collimating the light when suchsteep angles of incidence are employed.

Three bases set of 1024 theoretical SPR spectra fromangles of incidence between 56◦ and 58◦, 58◦ and 60◦, and60◦ and 62◦ were employed to model the specrum for a sam-ple RI of 1.34570 (5% KCl solution). In the model a fourlayer sapphire/chromium (1 nm)/gold (50 nm)/media (infi-nite) system was assumed. The complex permittivities ofgold, chromium, and sapphire were employed from literaturevalues[25] and interpolated based on a fifth order polynomial[1]. The non-negative-least squares optimization algorithmconverged to each basis set of spectra to fit the observed curve(Fig. 4). The distribution of the least squares fitting spectra(smoothing lines) is close to the measured spectral feature.The models show significant contribution in the observedSPR spectra from propagating modes as far away as 2◦ fromthe main SPR active mode. This effect is a direct result ofthe narrow distribution of modes propagating down the 0.12effective NA fiber and the inability to accurately and preciselypolish the sensing surface. For example, the fiber targeted top ct-i rw

esnele ructewt tweend rvedS ectra( (0,1 m1 y the

same theoretical and observed sensitivity (wavelength shiftper change in RI) over this range, but the standard deviation of50 replicate measurements for the 57◦/33◦ bare sapphire sen-sor (0.06–0.19 nm for the six samples) is significantly greaterthan the standard deviation for the 59◦/31◦ (0.02–0.08 nm)and the 61◦/29◦ (0.01–0.03 nm) configurations. The differ-ence in precision of estimating the wavelength position of theSPR can largely be explained by the relative widths of the SPRfeatures from the three sensors. The 57◦/33◦ configurationhas much broader dips due to the angular sensitivity on theincident photons; thus any movement in the fibers that effectchange in the distribution of light at the fiber jumper—sensorinterface is more likely to slightly change the observed SPRfeature.

The magnitude of difference between data points of SPRcoupling wavelength and their first-order polynomial fittingsis illustrated inFig. 6. The residuals of theoretical modelwere significantly less than that of the experiment, and theresiduals from experimental data were in±1 nm. In the prac-tical experiment, the reflection occurring at interfaces suchas the coupler|air|SPR probe will increase the backgroundlight intensity of the measured light, thus decreasing the SPRcoupling efficiency. Also, it is possible for the angular andlinear offset to mismatch between coupler and SPR probe,thus the SPR probes will not introduce parallel rays of light,which will not properly be coupled back into the coupler.

3

thescS lysis,n road,s atured d toc RTAt 0742

olish at 61◦ has its main contribution from a mode contang the SPR active surface at 62.8◦ incident from the fibealls.Based on the theoretical spectra predicted by the Fr

quations, three bare sapphire SPR probes were constith tapered tips of 57◦/33◦, 59◦/31◦, and 61◦/29◦, such

hat the sensors would be responsive to RI changes beistilled water and sea water (3.5% salinity). The obsePR features agree qualitatively with the predicted sp

Fig. 5). The probes were tested with six KCl solutions, 5, 10, 15, 20 wt.% KCl in distilled water) ranging fro.33315 to 1.35756 RI. All three probes have essentiall

d

.3. Fiber optic SPR probes in the hydrothermal fluids

Rapid thermal annealing (RTA) was found to improveharpness of the SPR spectral features.Fig. 7 shows typi-al observed SPR features for the 59◦/31◦ configuration for aapphire SPR sensor before and after RTA. For SPR anaarrow and deep features are considered preferable to bhallow spectral features. Thus, the aspect ratio of feepth divided by feature width at half depth can be useompare multiple methods of sensor preparation. Beforehe sapphire-fiber returned spectra with an aspect of 0.00

Y.-C. Kim et al. / Talanta 67 (2005) 908–917 913

Fig. 4. Least squares fitting SPR spectra due to incidence angle. Medium, 5% KCl solution in water (RI, 1.34570 at room temperature); Sapphire-fiber basedtapered probes: (1)θinc = 57◦; (2) θinc = 59◦; (3) θinc = 61◦; the atmosphere and room temperature.

Fig. 5. SPR spectra from tapered sapphire-fiber SPR probe (three angles and six RIs). Mediums, water (RI, 1.33315), 1% (RI, 1.33458), 5% (RI, 1.33956),10% (RI, 1.34571), 15% (RI, 1.35148), and 20% KCl (RI, 1.35756) solutions in water; Sapphire-fiber based tapered probes: (1) (θinc = 61◦); (2) (θinc = 59◦);(3) (θinc = 57◦); Kaiser Hollospec spectrograph with 449.61 nm of spectral coverage; the atmosphere and temperature, 20◦C.

914 Y.-C. Kim et al. / Talanta 67 (2005) 908–917

Fig. 6. Residuals of SPR coupling wavelength responses as a function of RI of mediums. Mediums, water (RI, 1.33315), 1% (RI, 1.33458), 5% (RI, 1.33956),10% (RI, 1.34571), 15% (RI, 1.35148), and 20% KCl (RI, 1.35756) solutions in water; Sapphire-fiber based tapered probes; (1) (θinc = 61◦); (2) (θinc = 59◦);(3) (θinc = 57◦); the atmosphere and temperature, 20◦C.

(dip/full width at half minimum) while after RTA the aspectimproved to 0.00106 (dip/full width at half minimum). Thisis an improvement of 29%. The improvement in SPR fea-ture characteristics might be caused by the relatively smoothsurface of gold and forming of shallow junction (Al2O3|Cr).

The sapphire sensors followed by RTA are capable ofmonitoring the RI of hydrothermal (high-temperature andpressure) fluids where silica-fibers fail. It is well know thatsilica is dissolved by hydrothermal water. Even coating thesensor with gold (inert to hydrothermal water) will not pre-vent the fiber degradation.Fig. 8d shows the silica-fiber basedSPR spectral feature from hydrothermal fluid at 3000 psi asthe temperature is increased from 220 to 260◦C. As thesilica-fiber degrades as a function of exposure time to dif-

F ium,w fiberbl spherea

ferent hydrothermal conditions, the gold flakes off and thecore becomes pitted and opaque. Consequently the reflectedlight decreases as the fiber degrades (Fig. 8c). A plot of SPRcoupling wavelengths versus pressure/temperature for theselected experiments is linear (Fig. 8a and b). As the temper-ature of samples is increased, the density decreases, leadingto a subsequent increase in the velocity of radiation and adecrease in the measured refractive index (i.e., SPR couplingwavelength). Also, the SPR coupling wavelength has red shiftas pressure increases, due to the resulting increase in density.The relation between refractive index (n) and density (ρ) ismost accurately given by the Lorentz–Lorenz expression:

RD

Mρ = n2 − 1

n2 + 2, (1)

where RD is the molar refraction andM is the molecularweight. The RD is nearly constant with changes in tem-perature and pressure by virtue of the density factor, whichis a function of pressure and temperature. Therefore, therefractive index is affected by changes in temperature, pres-sure, and sample’s concentration. From the Lorentz–Lorenzformular, the refractive index has linearity to the density.The linearity of the data (Fig. 8a and b) is consistent withthe Lorentz–Lorenz formular.

The sapphire-fiber based SPR probe, after the RTA,ws beene lingw phire-fi e seeni dictt raturec her-m and1 , theR s is3 ber

ig. 7. Effects of rapid thermal annealing (RTA) on SPR spectra. Medater; (1) before RTA treatment; (2) after RTA treatment; Sapphire-ased tapered probe (Cr, 3 nm + Au, 43 nm and (θinc = 59◦)); Kaiser Hol-

ospec spectrograph with 449.61 nm of spectral coverage; the atmond room temperature.

as exposed to hydrothermal water.Jobin-SPEX 270 Mpectrometer with 42.80 nm of spectral coverage hasmployed for these experiments. The plot of SPR coupavelengths versus pressure/temperature using the sapber based tapered SPR probe is also linear, as can bn Fig. 9a and b. Additionally, this data can be used to prehe changes of RI at some other pressure and tempeonditions. The RMSE of pressure calibration in hydrotal water, 5 and 10% KCl solutions is 255.90, 170.07,20.37 psi, respectively. In the temperature calibrationsMSE of hydrothermal water, 5 and 10% KCl solution.17, 4.89, and 8.34◦C, respectively. When the sapphire-fi

Y.-C. Kim et al. / Talanta 67 (2005) 908–917 915

Fig. 8. Silica-fiber based SPR spectra due to pressure and temperature. Silica-fiber based straight probe without RTA treatment. Medium, 5% KCl solutions inwater: (a) pressure effect at room temperature; (b) temperature effect at 3000 psi; (c) unprocessed reflection spectra as a function of temperature: (1) 220◦C;(2) 230◦C; (3) 240◦C; (4) 250◦C; (5) 260◦C; (d) normalized SPR spectra as a function of temperature: (1) 220◦C; (2) 230◦C; (3) 240◦C; (4) 250◦C; (5)260◦C; Kaiser Hollospec spectrograph with 449.61 nm of spectral coverage.

Fig. 9. Sapphire-fiber based SPR spectra due to pressure and temperature. Sapphire-fiber based tapered probe (θinc = 57◦) with RTA treatment. Mediums, water,5%, and 10% KCl solutions in water: (a) pressure effect at room temperature; (b) temperature effect at 3000 psi; (c) unprocessed reflection spectra as afunctionof temperature: (1) 220◦C; (2) 230◦C; (3) 240◦C; (4) 250◦C; (5) 260◦C; (d) normalized SPR spectra as a function of temperature: (1) 220◦C; (2) 230◦C; (3)240◦C; (4) 250◦C; (5) 260◦C; Jobin-SPEX 270 M spectrometer with 42.80 nm of spectral coverage.

916 Y.-C. Kim et al. / Talanta 67 (2005) 908–917

Fig. 10. Calibration curve at high pressure and temperature. Sapphire-fiberbased tapered probe (θinc = 57◦) after RTA treatment; thicker line, second-order polynomial fitting; mediums, 8%, 9%, 10%, 11%, 12%, 13%, and 14%KCl solutions in water; pressure, 3000 psi and temperature, 250◦C.

based SPR probes after the RTA were applied to monitor RIsof KCl solutions at each temperature in 3000 psi, the intensi-ties of reflection spectra were not loss until 260◦C (Fig. 9c).Consequently, the SPR dips were stable (Fig. 9d), and haveblue shift as temperature increases.

A quantitative study of sapphire-fiber based SPR sensor athigh-temperature (250◦C) and pressure (3000 psi) was per-formed. The calibration curve for seven KCl samples is shownin a plot sample’s concentration versus relative SPR couplingwavelength (Fig. 10). Each data point (circle) was calculatedby average of 50 SPR coupling wavelengths from each sam-ple. Examination of the calibration data shows that the experi-mental data points fit a curve (solid line) in the range of 8–14%KCl concentrations. An equation of the fitting curve is:

�λSPR= 0.0894�C2 − 0.935�C + 1.78, (2)

where�λSPR is the SPR coupling wavelength increase innm, �C is the KCl concentration increase in percentage.The RMSE of calibration across the range of 8.00–14.00%KCl solutions is 0.138% KCl. The error bounds (brokenlines) were produced by an interval of±RMSE.

The sapphire-fiber SPR sensor was stable throughout allthe experiments without loss of reflectivity or degradationof the SPR aspect. The stability of the sapphire was testedup to 3750 psi and 260◦C. Temperature above 260◦C wasn 202)e ttingf sa rmals ed. Int ondi-t 0f thes

4. Concluding remarks

A simple and rapid method for the evaluation of RIs inharsh environment was developed. This method is based onthe sapphire-fiber with SPR spectroscopy. The sapphire-fiberbased SPR sensor has advantages, such as short analysis timeand simplicity of use. Additionally, the applied sapphire-fiberbased SPR sensor could detect changes of SPR couplingwavelength at high pressure and temperature. In view of thesefacts, the sapphire-fiber based SPR sensors can be considereda valuable approach for in situ, real time RIs monitoring inthe harsh environment conditions.

Our current work is focusing on applying the rapid thermalannealing to sapphire-fiber SPR probes for the evaluation ofstability and sensitivity. This study is currently in progressand will be reported at a later date.

Acknowledgments

This research work was supported by NSF Oceans0119999. The authors thank Dr. Dong-Kyun Seo for pro-viding the rapid thermal annealing system.

References

ee,

m.

pt.

997)

a 62

ens.

[ K.S.

[ oin,

[ ie-04)

[ 7

[ ein-

[ 01)

[[ 997)

[[ ey,

net

ot tested because the epoxy (Tra-con, TRA-BOND Fmployed to mount the sapphire sensor in the swagelok fi

or the hydrothermal reactor is only rated to 265◦C. Pressurebove 3750 psi could not be maintained in the hydrotheystem; therefore more extreme systems were not testotal the sapphire-fiber was exposed to hydrothermal cions (arbitrarily defined as greater than 3000 psi and 20◦C)or more than 7 h during 14 days compared to 1 h forilica-fiber sensor.

[1] L.A. Obando, K.S. Booksh, Anal. Chem. 71 (1999) 5116.[2] L.S. Jung, C.T. Campbell, T.M. Chinowsky, M.N. Mar, S.S. Y

Langmuir 14 (1998) 5636.[3] M.D. Malinsky, K.L. Kelly, G.C. Schatz, R.P.V. Duyne, J. A

Chem. Soc. 123 (2001) 1471.[4] H.E.d. Bruijn, B.S.F. Altenberg, R.P.H. Kooyman, J. Greve, O

Commun. 82 (1991) 425.[5] E. Kretschmann, Zeitschrift fuer Physik 221 (1969) 346.[6] E. Kretschmann, Zeitschrift fuer Physik 221 (1969) 357.[7] K.S. Johnston, S.S. Yee, K.S. Booksh, Anal. Chem. 69 (1

1844.[8] J.-F. Masson, L.A. Obando, S. Beaudoin, K.S. Booksh, Talant

(2004) 865.[9] L.A. Obando, D.J. Gentleman, J.R. Holloway, K.S. Booksh, S

Actuators B 100 (2004) 439.10] D.J. Gentleman, L.A. Obando, J.-F. Masson, J.R. Holloway,

Booksh, Anal. Chim. Acta 515 (2004) 291.11] J.-F. Masson, T.M. Battaglia, Y.-C. Kim, A. Prakash, S. Beaud

K.S. Booksh, Talanta 64 (2004) 716.12] J.-F. Masson, M. Barnhart, T.M. Battaglia, G.E. Morris, R.A. N

man, P.J. Young, C.L. Lorson, K.S. Booksh, Analyst 129 (20855.

13] J.-F. Masson, T. Hartmann, H. Durr, K.S. Booksh, Opt. Mater. 2(2004) 435.

14] M. Niggemann, A. Katerkamp, M. Pellmann, P. Bolsmann, J. Rbold, K. Cammann, Sens. Actuators B 34 (1996) 328.

15] A. Diez, M.V. Andres, J.L. Cruz, Sens. Actuators B 73 (2095.

16] R.C. Jorgenson, Sens. Actuators B 73 (2001) 236.17] C.P. Cahill, K.S. Johnston, S.S. Yee, Sens. Actuators B 45 (1

161.18] R.C. Jorgenson, S.S. Yee, Sens. Actuators A 45 (1994) 44.19] D.J. Fornari, T. Shank, K.L. Von Damm, T.K.P. Gregg, M. Lill

G. Levai, A. Bray, R.M. Haymon, M.R. Perfit, R. Lutz, Earth PlaSci. Lett. 160 (1998) 419.

Y.-C. Kim et al. / Talanta 67 (2005) 908–917 917

[20] K. Ding, W.E. Seyfried, M.K. Tivey, A.M. Bradley, Earth Planet Sci.Lett. 186 (2001) 417.

[21] T.F. Rozan, S.M. Theberge, G. Luther, Anal. Chim. Acta 415 (2000)175.

[22] A. Ishimaru, Electromagnetic Wave Propagation, Radiation, andScattering, Prentice-Hall, New Jersey, 1991 (Chapter 3).

[23] K.S. Johnston, S.R. Karlsen, C.C. Jung, S.S. Yee, Mater. Chem.Phys. 42 (1995) 242.

[24] R.K. Nubling, J.A. Harrington, Appl. Opt. 36 (1997) 5934.[25] E.D. Palik, Handbook of Optical Constants of Solids, Academic

Press, New York, 1985.